type: runbook
status: active
timestamp: 2026-06-22
tags: [runbook, hosting, free-tier, queues, pubsub, cloudflare-queues, upstash-qstash, inngest]
status: active
timestamp: 2026-06-22
tags: [runbook, hosting, free-tier, queues, pubsub, cloudflare-queues, upstash-qstash, inngest]
Free hosting — queues + pub-sub (CF Queues, Upstash QStash, Inngest, Trigger.dev, Pusher)
Queues + pub-sub free tiers: CF, QStash, Inngest
Queues + pub-sub — free tiers (2026-06-22)
Until Feb 4 2026, Cloudflare Queues required the Workers Paid plan. Now it’s in the Workers Free tier with a 10K ops/day cap — the biggest 2026 unlock in this category.
The table
| # | Provider | Free tier | Card@signup | Card to use free | KYC | Verdict |
|---|---|---|---|---|---|---|
| 1 | Cloudflare Queues ⭐ | 10,000 ops/day on Workers Free (GA in free 2026-02-04), 24-hour retention | NO | NO | NO | KEEP |
| 2 | Upstash QStash ⭐ | 1,000 messages/day, 10 active schedules, 50 GB bandwidth, 1 MB msg, retries free | NO | NO | NO | KEEP |
| 3 | Inngest | 50,000 executions/mo, 100K events, 5 concurrent steps, 50 realtime conns, 24h traces | NO | NO | NO | KEEP |
| 4 | Trigger.dev | ~10,000 runs/mo, limited concurrency, 14-day history, 1 prod env. Apache 2.0 self-host = unlimited | NO | NO | NO | KEEP |
| 5 | Pusher Channels Sandbox | 200,000 messages/day, 100 concurrent connections | NO | NO | NO | KEEP |
How the family uses queues
Today: very lightly. Most async work fits one of:
- Cron Triggers (Cloudflare Workers scheduled functions) — daily/hourly tasks
- GitHub Actions scheduled workflows — multi-step, depend-on-secrets tasks
- Synchronous HTTP — most user-triggered work fits in the 10–50 ms CPU budget
When this is not enough (long-running webhook processing, fan-out work, durable retries):
- Cloudflare Queues for producer → consumer Worker pipelines (now free!)
- Upstash QStash for “I have an HTTP endpoint and want it called with retry/delay/cron”
- Inngest for step-functions-style multi-step durable workflows
- Trigger.dev if we want self-hostable + Apache 2.0 (no vendor risk)
Quirks per provider
- Cloudflare Queues 10K ops/day. 1 op = produce + 1 op per consumer. For burst-tolerant scheduling this is plenty; for high-traffic event pipelines it’s the wall.
- Upstash QStash 1K msg/day + 10 schedules. “10 active schedules” is the under-the-radar cap — for cron-replacement use cases, 10 named schedules across the family is the real ceiling.
- Inngest free is the most generous step-function plan in the table. 50K executions/mo + 100K events/mo + 24h traces. Steps are the granularity (so a 5-step workflow = 5 executions).
- Trigger.dev self-host trump card. Apache 2.0, you can run it on Render Free or Koyeb Free and never pay them. Best fit if we ever build a long-running scheduling system.
- Pusher Channels Sandbox. Realtime websockets, not really a queue. 200K msgs/day is fine for chat/presence in a single app, not 50 apps.
Recommendation for the family
- Pub-sub for Worker pipelines: Cloudflare Queues (now free).
- HTTP cron / delay / retry: Upstash QStash (10 schedules covers most of what we’d need).
- Multi-step durable workflows: Inngest free tier.
- Reach for self-host: Trigger.dev on Render Free, only if vendor risk worries us.
- Realtime websockets: Pusher Sandbox for one app; Cloudflare Durable Objects for fleet-wide.
Sources
- Cloudflare Queues pricing — Feb 2026 GA in free
- Upstash QStash pricing — 1K msgs/day
- Inngest pricing — 50K execs/mo
- Trigger.dev pricing — free + self-host
- Pusher pricing — Sandbox tier